diff --git a/ui/app/(prowler)/lighthouse/_lib/model-loading.test.ts b/ui/app/(prowler)/lighthouse/_lib/model-loading.test.ts
new file mode 100644
index 0000000000..7298b7243a
--- /dev/null
+++ b/ui/app/(prowler)/lighthouse/_lib/model-loading.test.ts
@@ -0,0 +1,126 @@
+import { describe, expect, it, vi } from "vitest";
+
+import type {
+  LighthouseV2Configuration,
+  LighthouseV2ProviderType,
+  LighthouseV2SupportedModel,
+} from "@/app/(prowler)/lighthouse/_types";
+
+import { loadLighthouseV2ConnectedModels } from "./model-loading";
+
+describe("loadLighthouseV2ConnectedModels", () => {
+  it("loads models only for connected configurations and initializes the rest as empty", async () => {
+    // Given
+    const openAIModel = model("gpt-5.5");
+    const loadModels = vi.fn(async (providerType: LighthouseV2ProviderType) => {
+      if (providerType === "openai-compatible") {
+        return { error: "Connection failed", status: 400 };
+      }
+
+      return { data: [openAIModel] };
+    });
+
+    // When
+    const result = await loadLighthouseV2ConnectedModels(
+      [
+        configuration("openai", true),
+        configuration("bedrock", false),
+        configuration("openai-compatible", false),
+      ],
+      loadModels,
+    );
+
+    // Then
+    expect(loadModels).toHaveBeenCalledTimes(1);
+    expect(loadModels).toHaveBeenCalledWith("openai");
+    expect(result.modelsByProvider).toEqual({
+      openai: [openAIModel],
+      bedrock: [],
+      "openai-compatible": [],
+    });
+    expect(result.failedModelProviders).toEqual([]);
+  });
+
+  it("reports model-loading failures only for connected configurations", async () => {
+    // Given
+    const loadModels = vi.fn(async (providerType: LighthouseV2ProviderType) => {
+      if (providerType === "bedrock") {
+        return { error: "Bedrock models unavailable", status: 503 };
+      }
+
+      throw new Error(`Disconnected provider should not load: ${providerType}`);
+    });
+
+    // When
+    const result = await loadLighthouseV2ConnectedModels(
+      [
+        configuration("openai", false),
+        configuration("bedrock", true),
+        configuration("openai-compatible", false),
+      ],
+      loadModels,
+    );
+
+    // Then
+    expect(loadModels).toHaveBeenCalledTimes(1);
+    expect(loadModels).toHaveBeenCalledWith("bedrock");
+    expect(result.modelsByProvider).toEqual({
+      openai: [],
+      bedrock: [],
+      "openai-compatible": [],
+    });
+    expect(result.failedModelProviders).toEqual(["bedrock"]);
+  });
+
+  it("dedupes provider types and treats null connection state as disconnected", async () => {
+    // Given
+    const openAIModel = model("gpt-5.5");
+    const loadModels = vi.fn(async () => ({ data: [openAIModel] }));
+
+    // When
+    const result = await loadLighthouseV2ConnectedModels(
+      [
+        configuration("openai", true),
+        { ...configuration("openai", true), id: "config-openai-2" },
+        { ...configuration("bedrock", false), connected: null },
+      ],
+      loadModels,
+    );
+
+    // Then
+    expect(loadModels).toHaveBeenCalledTimes(1);
+    expect(loadModels).toHaveBeenCalledWith("openai");
+    expect(result.modelsByProvider.openai).toEqual([openAIModel]);
+    expect(result.modelsByProvider.bedrock).toEqual([]);
+  });
+});
+
+function configuration(
+  providerType: LighthouseV2ProviderType,
+  connected: boolean,
+): LighthouseV2Configuration {
+  return {
+    id: `config-${providerType}`,
+    providerType,
+    baseUrl:
+      providerType === "openai-compatible" ? "https://example.com" : null,
+    defaultModel: null,
+    businessContext: "Production account",
+    connected,
+    connectionLastCheckedAt: null,
+    insertedAt: "2026-06-24T09:00:00Z",
+    updatedAt: "2026-06-24T10:00:00Z",
+  };
+}
+
+function model(id: string): LighthouseV2SupportedModel {
+  return {
+    id,
+    name: id,
+    maxInputTokens: null,
+    maxOutputTokens: null,
+    supportsFunctionCalling: true,
+    supportsVision: false,
+    supportsReasoning: true,
+  };
+}
diff --git a/ui/app/(prowler)/lighthouse/_lib/model-loading.ts b/ui/app/(prowler)/lighthouse/_lib/model-loading.ts
new file mode 100644
index 0000000000..cbac91559e
--- /dev/null
+++ b/ui/app/(prowler)/lighthouse/_lib/model-loading.ts
@@ -0,0 +1,79 @@
+import {
+  LIGHTHOUSE_V2_PROVIDER_TYPE,
+  type LighthouseV2Configuration,
+  type LighthouseV2ProviderType,
+  type LighthouseV2SupportedModel,
+} from "@/app/(prowler)/lighthouse/_types";
+
+interface LighthouseV2SupportedModelsSuccess {
+  data: LighthouseV2SupportedModel[];
+}
+
+interface LighthouseV2SupportedModelsFailure {
+  error: string;
+  errors?: unknown[];
+  status?: number;
+}
+
+type LighthouseV2SupportedModelsResult =
+  | LighthouseV2SupportedModelsSuccess
+  | LighthouseV2SupportedModelsFailure;
+
+type LoadLighthouseV2SupportedModels = (
+  providerType: LighthouseV2ProviderType,
+) => Promise<LighthouseV2SupportedModelsResult>;
+
+interface LighthouseV2ConnectedModelsResult {
+  modelsByProvider: Record<
+    LighthouseV2ProviderType,
+    LighthouseV2SupportedModel[]
+  >;
+  failedModelProviders: LighthouseV2ProviderType[];
+}
+
+export function createEmptyLighthouseV2ModelsByProvider(): Record<
+  LighthouseV2ProviderType,
+  LighthouseV2SupportedModel[]
+> {
+  return {
+    [LIGHTHOUSE_V2_PROVIDER_TYPE.OPENAI]: [],
+    [LIGHTHOUSE_V2_PROVIDER_TYPE.BEDROCK]: [],
+    [LIGHTHOUSE_V2_PROVIDER_TYPE.OPENAI_COMPATIBLE]: [],
+  };
+}
+
+export async function loadLighthouseV2ConnectedModels(
+  configurations: LighthouseV2Configuration[],
+  loadModels: LoadLighthouseV2SupportedModels,
+): Promise<LighthouseV2ConnectedModelsResult> {
+  const modelsByProvider = createEmptyLighthouseV2ModelsByProvider();
+  // Disconnected providers keep the [] pre-seed and never hit the models
+  // endpoint, so they can't surface spurious load failures either.
+  const connectedProviderTypes = Array.from(
+    new Set(
+      configurations
+        .filter((configuration) => configuration.connected === true)
+        .map((configuration) => configuration.providerType),
+    ),
+  );
+
+  const modelsEntries = await Promise.all(
+    connectedProviderTypes.map(async (providerType) => {
+      const result = await loadModels(providerType);
+      return [providerType, result] as const;
+    }),
+  );
+
+  const failedModelProviders: LighthouseV2ProviderType[] = [];
+
+  modelsEntries.forEach(([providerType, result]) => {
+    if ("data" in result) {
+      modelsByProvider[providerType] = result.data;
+      return;
+    }
+
+    failedModelProviders.push(providerType);
+  });
+
+  return { modelsByProvider, failedModelProviders };
+}
diff --git a/ui/app/(prowler)/lighthouse/page.tsx b/ui/app/(prowler)/lighthouse/page.tsx
index 59b28ced43..0724e2fa45 100644
--- a/ui/app/(prowler)/lighthouse/page.tsx
+++ b/ui/app/(prowler)/lighthouse/page.tsx
@@ -13,11 +13,8 @@ import {
 } from "@/app/(prowler)/lighthouse/_actions";
 import { LighthouseV2ChatPage } from "@/app/(prowler)/lighthouse/_components/chat";
 import { LighthouseV2NavigationModeSync } from "@/app/(prowler)/lighthouse/_components/navigation";
+import { loadLighthouseV2ConnectedModels } from "@/app/(prowler)/lighthouse/_lib/model-loading";
 import { buildLighthouseV2StreamUrl } from "@/app/(prowler)/lighthouse/_lib/stream-url";
-import type {
-  LighthouseV2ProviderType,
-  LighthouseV2SupportedModel,
-} from "@/app/(prowler)/lighthouse/_types";
 import { LighthouseIcon } from "@/components/icons/Icons";
 import { Chat } from "@/components/lighthouse-v1";
 import { ContentLayout } from "@/components/ui";
@@ -53,25 +50,14 @@ export default async function AIChatbot({
       return redirect("/lighthouse/settings");
     }
 
-    const modelsEntries = await Promise.all(
-      configurations.map(async (configuration) => {
-        const result = await getLighthouseV2SupportedModels(
-          configuration.providerType,
-        );
-        return [configuration.providerType, result] as const;
-      }),
-    );
-    const modelsByProvider = Object.fromEntries(
-      modelsEntries.map(([providerType, result]) => [
-        providerType,
-        "data" in result ? result.data : [],
-      ]),
-    ) as Record<LighthouseV2ProviderType, LighthouseV2SupportedModel[]>;
-    // Surface (rather than silently swallow to []) providers whose models
-    // failed to load, so an empty model list reads as a real backend failure.
-    const failedModelProviders = modelsEntries
-      .filter(([, result]) => !("data" in result))
-      .map(([providerType]) => providerType);
+    const { modelsByProvider, failedModelProviders } =
+      await loadLighthouseV2ConnectedModels(
+        configurations,
+        getLighthouseV2SupportedModels,
+      );
+    // Surface (rather than silently swallow to []) connected providers whose
+    // models failed to load, so their empty list reads as a real backend
+    // failure. Disconnected providers are never fetched (see model-loading.ts).
     const modelsError =
       failedModelProviders.length > 0
         ? `Could not load available models for: ${failedModelProviders.join(", ")}. Try again shortly.`